Merge branch 'master' into cuda_a_dev

Common/helper_cuda.h

@@ -138,26 +138,26 @@ static const char *_cudaGetErrorEnum(cufftResult error) {
     case CUFFT_UNALIGNED_DATA:
       return "CUFFT_UNALIGNED_DATA";
 
-    case CUFFT_INCOMPLETE_PARAMETER_LIST:
-      return "CUFFT_INCOMPLETE_PARAMETER_LIST";
-
     case CUFFT_INVALID_DEVICE:
       return "CUFFT_INVALID_DEVICE";
 
-    case CUFFT_PARSE_ERROR:
-      return "CUFFT_PARSE_ERROR";
-
     case CUFFT_NO_WORKSPACE:
       return "CUFFT_NO_WORKSPACE";
 
     case CUFFT_NOT_IMPLEMENTED:
       return "CUFFT_NOT_IMPLEMENTED";
 
-    case CUFFT_LICENSE_ERROR:
+    case CUFFT_MISSING_DEPENDENCY:
-      return "CUFFT_LICENSE_ERROR";
+      return "CUFFT_MISSING_DEPENDENCY";
 
-    case CUFFT_NOT_SUPPORTED:
+    case CUFFT_NVRTC_FAILURE:
-      return "CUFFT_NOT_SUPPORTED";
+      return "CUFFT_NVRTC_FAILURE";
+
+    case CUFFT_NVJITLINK_FAILURE:
+      return "CUFFT_NVJITLINK_FAILURE";
+
+    case CUFFT_NVSHMEM_FAILURE:
+      return "CUFFT_NVSHMEM_FAILURE";
   }
 
   return "<unknown>";

Samples/0_Introduction/matrixMul_nvrtc/CMakeLists.txt

@@ -45,10 +45,10 @@ add_custom_command(TARGET matrixMul_nvrtc POST_BUILD
 
 add_custom_command(TARGET matrixMul_nvrtc POST_BUILD
     COMMAND ${CMAKE_COMMAND} -E copy_directory
-    ${CUDAToolkit_BIN_DIR}/../include/nv ${CMAKE_CURRENT_BINARY_DIR}/nv
+    ${CUDAToolkit_BIN_DIR}/../include/cccl/nv ${CMAKE_CURRENT_BINARY_DIR}/nv
 )
 
 add_custom_command(TARGET matrixMul_nvrtc POST_BUILD
     COMMAND ${CMAKE_COMMAND} -E copy_directory
-    ${CUDAToolkit_BIN_DIR}/../include/cuda ${CMAKE_CURRENT_BINARY_DIR}/cuda
+    ${CUDAToolkit_BIN_DIR}/../include/cccl/cuda ${CMAKE_CURRENT_BINARY_DIR}/cuda
 )

Samples/4_CUDA_Libraries/conjugateGradientMultiBlockCG/CMakeLists.txt

@@ -13,6 +13,7 @@ set(CMAKE_CUDA_FLAGS "${CMAKE_CUDA_FLAGS} -Wno-deprecated-gpu-targets")
 
 if(ENABLE_CUDA_DEBUG)
     set(CMAKE_CUDA_FLAGS "${CMAKE_CUDA_FLAGS} -G")        # enable cuda-gdb (may significantly affect performance on some targets)
+    set(CMAKE_CUDA_FLAGS "${CMAKE_CUDA_FLAGS} -maxrregcount=128") # limit register usage to 128 per thread to comply with the maximum number of 32-bit registers per SM
 else()
     set(CMAKE_CUDA_FLAGS "${CMAKE_CUDA_FLAGS} -lineinfo") # add line information to all builds for debug tools (exclusive to -G option)
 endif()

Samples/5_Domain_Specific/simpleVulkan/VulkanBaseApp.h

@@ -34,8 +34,10 @@
 #include <vulkan/vulkan.h>
 #ifdef _WIN64
 #define NOMINMAX
-#include <vulkan/vulkan_win32.h>
+// Add windows.h to the include path
 #include <windows.h>
+// Add vulkan_win32.h to the include path
+#include <vulkan/vulkan_win32.h>
 #endif /* _WIN64 */
 
 /* remove _VK_TIMELINE_SEMAPHORE to use binary semaphores */

Samples/5_Domain_Specific/simpleVulkanMMAP/VulkanBaseApp.h

@@ -34,8 +34,10 @@
 #include <vulkan/vulkan.h>
 #ifdef _WIN64
 #define NOMINMAX
-#include <vulkan/vulkan_win32.h>
+// Add windows.h to the include path firstly as dependency for other Windows headers
 #include <windows.h>
+// Add other Windows headers
+#include <vulkan/vulkan_win32.h>
 #endif /* _WIN64 */
 
 struct GLFWwindow;

Samples/5_Domain_Specific/vulkanImageCUDA/vulkanImageCUDA.cu

@@ -27,10 +27,12 @@
 
 #define GLFW_INCLUDE_VULKAN
 #ifdef _WIN64
+// Add windows.h to the include path firstly as dependency for other Windows headers
+#include <windows.h>
+// Add other Windows headers
 #include <VersionHelpers.h>
 #include <aclapi.h>
 #include <dxgi1_2.h>
-#include <windows.h>
 #define _USE_MATH_DEFINES
 #endif
 

Samples/6_Performance/transpose/transpose.cu

@@ -606,8 +606,7 @@ int main(int argc, char **argv)
             h_odata[i] = 0;
         }
         // copy host data to device
-        checkCudaErrors(
+        checkCudaErrors(cudaMemcpy(d_odata, h_odata, mem_size, cudaMemcpyHostToDevice));
-            cudaMemcpy(d_odata, h_odata, mem_size, cudaMemcpyHostToDevice));
     }
 
     // cleanup

Samples/8_Platform_Specific/Tegra/CMakeLists.txt

@@ -5,7 +5,7 @@ add_subdirectory(cuDLAHybridMode)
 add_subdirectory(cuDLALayerwiseStatsHybrid)
 add_subdirectory(cuDLALayerwiseStatsStandalone)
 add_subdirectory(cuDLAStandaloneMode)
-#add_subdirectory(EGLSync_CUDAEvent_Interop)
+add_subdirectory(EGLSync_CUDAEvent_Interop)
 add_subdirectory(fluidsGLES)
 add_subdirectory(nbody_opengles)
 add_subdirectory(simpleGLES)

Samples/8_Platform_Specific/Tegra/EGLSync_CUDAEvent_Interop/CMakeLists.txt

@@ -17,7 +17,7 @@ else()
 endif()
 
 # Include directories and libraries
-include_directories(../../../Common)
+include_directories(../../../../Common)
 
 find_package(EGL)
 find_package(X11)

Samples/8_Platform_Specific/Tegra/EGLSync_CUDAEvent_Interop/EGLSync_CUDAEvent_Interop.cu

@@ -259,7 +259,8 @@ void checkSync(int argc, char **argv)
     unsigned char *pSurf_read = NULL, *pSurf_write = NULL;
     int            integrated;
 
-    CUresult status = CUDA_SUCCESS;
+    CUresult          status          = CUDA_SUCCESS;
+    CUctxCreateParams ctxCreateParams = {};
 
     // Init values for variables
     x = y = 0;
@@ -269,7 +270,7 @@ void checkSync(int argc, char **argv)
     }
     device = findCudaDeviceDRV(argc, (const char **)argv);
 
-    if (CUDA_SUCCESS != (status = cuCtxCreate(&context, 0, device))) {
+    if (CUDA_SUCCESS != (status = cuCtxCreate(&context, &ctxCreateParams, 0, device))) {
         printf("failed to create CUDA context\n");
     }
     cuCtxPushCurrent(context);